How we measure robot training pay

• We pull each employer's job feed every day and read the hourly rates off it, then take the median and the typical pay across the open roles that list a rate. Typical pay runs from the 25th percentile of what roles pay (the middle of each advertised range) to the 90th-percentile advertised ceiling, so the better-paid on-site operator and vehicle-operator roles fall inside it. It is worked out the same way for every list of roles (the whole board, one employer, one kind of work, one employer's roles of one kind, or a search), and a list needs at least 3 distinct rates before it shows one. Each pay figure opens the roles it was measured from, and that list shows the same figure. No figure includes salaries or rates paid per task or per footage.
• Duplicate relistings of one job are removed before we count distinct advertised rates, so a role posted over and over cannot pull the median its way. A rate count and an open-role count measure different things, and we never pass one off as the other.
• Advertised ceilings ("up to $X/hr") are left out: they are marketing maximums, not wages anyone was seen earning. If an employer's feed gives nothing but ceilings, it gets no live median, and the page says so instead of passing a ceiling off as the going rate.
• A figure earns the name market rate only when it covers three or more employers and a big enough sample. A rate from one employer appears as that employer's rate card, labeled as such, and is never published as the going rate for the category. Breadth does not equal representativeness: missing pay, geography and role mix can still skew a sample, and the note beneath each live band states what it was measured from.

Some links here are referral links: when you join an employer through one, RobotTraining.jobs may receive a commission. It costs you nothing, and the employer pays you exactly the same.

Referral fees are only part of how we earn. We also take on paid expert-data projects: a company building robots or AI pays us for a data job, and professionals from our expert bench do the work under our expert terms, paid by us. None of that feeds a figure or an order described on this page: a client, current or possible, is never moved up a list, and its pay is compiled exactly like any other employer's. Some employers listed here may already be clients or become ones, and a few sell the same kind of projects themselves.

Below is precisely where referral ties shape what you see, and where they do not. The facts are off limits: a pay figure and its confidence label are put together identically whether an employer pays us or not. The two sort options you can pick, highest pay and newest first, order roles only by the number and the date, and referrals play no part.

On the home page, the Top Networks list is a featured list. It holds the programs we have a referral relationship with, ordered by what each visit we send there is worth to us: the share of the people we send who end up paid there, multiplied by the referral fee, taken from each program's own referral records and updated whenever they change. The line directly beneath it states that joining through those links may earn us a referral fee. The home page's employer filter and its “Employers hiring right now” list follow the single order every list of employers uses, set out below, and the referral disclosure on that page applies to both.

Every other list of employers is ordered the same way: the groups in the full employer directory, the employers hiring for a kind of work or in a country, the alternatives to an employer and the employers to compare it with, the comparison pages, the employer guides in Portuguese and Spanish, the employers on the page about how the site works, and the employers suggested in your account. Each list sorts by what applying to an employer is worth, explained in the next section, multiplied by one plus a referral boost for the programs we have a referral relationship with. The boost scales in proportion, from 20% for the program whose visits are worth least to us up to 100% for the one worth most, so a partner gets a lift but is never placed first whatever its worth. There is no boost while a program cannot credit new sign-ups (its referral cap is reached), when you already work there, or when it is closing to new people. Employers of equal worth keep the list's own order (its open-role count, where the list shows one), and a sentence naming the employers with the most roles names them by role count alone. Lists of roles, the similar roles on a job page among them, give partner roles the same boost inside the one figure that orders every list of roles (further down). The pay figures, and the highest-pay and newest sorts, never carry any referral input.

One figure orders every list of roles here: the job board, the AI search and your matches, the roles of an employer, a kind of work or a country, the similar roles shown beside a role, and the roles in our emails and messages. It multiplies four things: how likely the role suits you, how strongly people respond to it for the attention it has had, how crowded it already is, and one plus the referral boost of an employer we have a referral relationship with. How likely it suits you is our AI check's verdict where it has checked that role for you (a strong or possible match suits you, an unlikely one does not); otherwise it is the chance a model estimates from how close the role sits to your profile, trained on every check our AI has made. A list made for nobody in particular treats every role alike on this point. How strongly people respond is the role's apply clicks per real view (a role is seen once at least half of it stays on screen for a second), borrowing from the roles nearest to it in meaning while few people have seen it, so a role rises or falls with what people do when they see it, not with its age or the slot a list gave it. An employer takes only so many people for a role, so crowding counts how many of our visitors already applied and how much that cuts the next applicant's chance of being picked: the chance with that many people ahead divided by the chance with nobody ahead, measured on the applications whose outcome people reported to us, each employer leaning on the average employer while it has few outcomes. A role nobody has applied to yet counts in full, and a similar role with fewer people ahead moves up, which spreads applicants toward the roles where each stands a better chance. The referral boost runs from 20% to 100%, by what a visit to that employer is worth to us, and is zero while it cannot credit new sign-ups, at an employer you already work with, and at any employer without a referral relationship. A partner's role therefore weighs more, yet a role that suits you better, or that far more people want, still ranks above it, and partner roles land wherever that figure places them. When a role truly pays above the market it carries that fact as a tag (“Top-tier pay” for the top quarter of what open roles pay, “Above-typical pay” for more than the typical role), measured against the same live rates as the rest of the site, so the tag reflects something real about the role rather than a label. Sorting by Highest pay or Newest puts the number or the date first.

Describe the job you want, or match against your LinkedIn profile, and an AI model reads your request. A role counts as a match when it is more likely to suit you than not, or when its title contains a word you typed. Matches come first and every other role follows, both in the same figure. Under Newest or Top pay, matching roles are ordered by the date or the number first.

Roles similar to a role use that same figure, except that how likely one suits you becomes how likely a person suited to the first role suits this one. A model estimates that from every pair of roles our AI check has judged for the same person (how close the two are in meaning, the language and field their titles name, the kind of work and the employer). As in the AI search, the roles more likely to suit than not come first, each placed by the figure, with the rest after them, so a role many people want never jumps ahead of a closer one that suits. This order is used for the list on a role's page, the open roles next to a role that has closed, the partner roles in a role's summary and below the roles of an employer we have no referral relationship with, and “Similar roles hiring now”, a small card that appears after you apply at such an employer. That card shows at most three roles at other employers, open in your location, never one you were already shown, and never during a visit in which you already applied through a referral link.

Click a role and a brief summary opens: the pay, where the role is open, the first part of its description, and what kind of employer it is. If you have a free profile, an AI model also compares the role with your background and says whether you look like a strong, possible or unlikely match, with its reasons. Roles at our referral partners show their apply link in the summary right away. A role at an employer we have no referral relationship with has no apply link in the summary, with or without a profile: the summary offers to check whether you qualify, and the role's own job page links to the employer's posting. Next to a role like that, the summary offers up to three of the roles closest to it at our referral partners that you can apply to immediately, open where you are and listed with their pay and how each one pays; once our AI has checked them for you, the ones that suit you come first.

Our emails and messages use the same figure too: the new-role alerts for experts, the roles our assistant suggests, and our Telegram channel with its alerts. The weekly roles email and the welcome email show only roles first posted that week, ordered by the figure, because they promise what is new. An invitation to refer a colleague shows roles at employers that pay us a referral fee, the ones that would earn you most first. Each of these says so. Apps that read our roles from inside an AI assistant or an app directory (our MCP server, the jobs API, Poe) receive a neutral order, as those directories require.

In every case, this referral disclosure shows up wherever partner status changes the order, and any link you click sends you to the employer on that employer's own terms.